OISST Sea Surface Temperature Report
Adam A. Kemberling
Updated on: 2021-04-19
Caribbean Sea Sea Surface Temperature Outlook
Region Extent
The spatial extent for Caribbean Sea is displayed below. This bounding box is the same bounding box coordinates used to clip the OISST data when constructing the time series data from the array.
# testing regions
# params$region <- "agulhas_current"
# File paths for various extents based on params$region
region_paths <- get_timeseries_paths(region_group = "lme")
# Load the bounding box for Andy's GOM to show they align
poly_path <- region_paths[[params$region]][["shape_path"]]
region_extent <- st_read(poly_path, quiet = TRUE)
# Pull extents for the region for crop
crop_x <- st_bbox(region_extent)[c(1,3)]
crop_y <- st_bbox(region_extent)[c(2,4)]
# Zoom out for cpr extent
if(tolower(params$region) == "cpr gulf of maine"){
crop_x <- c(-70.875, -65.375)
crop_y <- c(40.375, 45.125)}
# one off shapes
# region_extent <- read_sf(paste0(res_path, ""))
# Full plot
ggplot() +
geom_sf(data = world, fill = "gray90") +
geom_sf(data = region_extent,
color = gmri_cols("orange"),
fill = gmri_cols("orange"),
alpha = 0.2,
linetype = 2,
size = 0.5) +
map_theme 
OISSTv2 Resources on Box
For many of the area we look at frequently here at GMRI, regional climatologies and anomalies have been processed for quick access to aid in other research endeavors.
These resources were processed using a handful of jupyter notebook work flows to take advantage of {xarray}.
The following sections will highlight what those work flows are, what they detail, and how they can be updated using the one of our focal regions as an example.
The root folder for the following OISST products is ~/Box/RES_Data/OISST/oisst_mainstays. Resources can be accessed manually through the box folder or more directly using the gmRi R package.
Regional Timeseries
Area-specific time series are the most basic building block for relaying temporal trends. For any desired area (represented by a spatial polygon) we can generate a time series table of the mean sea surface temperature within that area for each day. Additionally, we can compare how observed temperatures correspond with the expected conditions based on a climatology using a specified reference period.
Accessing the Data
Pre-processed tables have been placed on box for quick access from within the RES_Data folder: RES_Data/OISST/oisst_mainstays/regional_timeseries.
A function to quickly access these time-lines has been added to the {gmRi} package as oisst_access_timeseries(). Additional functions exist to look up what timeseries are availablegmRi::get_region_names(). And an additional function was made to return both the path to the timeseries data and the path to the shapefile used to generate the timeseries using: gmRi::get_timeseries_paths().
# Use {gmRi} instead to load timeseries to tie up loose ends
timeseries_path <- region_paths[[params$region]][["timeseries_path"]]
region_timeseries <- read_csv(timeseries_path, col_types = cols())
# format dates
region_timeseries <- region_timeseries %>%
mutate(time = as.Date(time))
# Display Table of first 6 entries
head(region_timeseries) %>%
mutate(across(where(is.numeric), round, 2)) %>%
kable() %>%
kable_styling()| time | sst | modified_ordinal_day | sst_clim | clim_sd | sst_anom |
|---|---|---|---|---|---|
| 1981-09-01 | 8.58 | 245 | 9.44 | 0.75 | -0.86 |
| 1981-09-02 | 8.75 | 246 | 9.45 | 0.74 | -0.70 |
| 1981-09-03 | 8.85 | 247 | 9.44 | 0.74 | -0.59 |
| 1981-09-04 | 9.37 | 248 | 9.44 | 0.74 | -0.07 |
| 1981-09-05 | 9.42 | 249 | 9.44 | 0.74 | -0.01 |
| 1981-09-06 | 9.77 | 250 | 9.41 | 0.74 | 0.36 |
# march 1st sst
mar1 <- region_timeseries %>%
filter(modified_ordinal_day == 61) %>%
distinct(sst_clim) %>%
pull(sst_clim)Each of our Climatologies are currently set up to calculate daily averages on a modified year day, such that every March 1st and all days after fall on the same day, regardless of whether it is a leap year or not.
This preserves comparisons across calendar dates such-as: “The average temperature on march 1st is 1.7556444` for the reference period 1982 to 2011”
In these tables sst is the mean temperature observed for that date averaged across all cells within the area. sst_clim & clim_sd are the climate means and standard deviations for a 1982-2011 climatology. sst_anom is the daily observed minus the climate mean.
Warming Rates
Regional warming trends below were calculated using all the available data for complete years beginning with 1982 through the end of 2020.
Annual
# Summarize by year to return mean annual anomalies and variance
annual_summary <- region_timeseries %>%
mutate(year = year(time)) %>%
group_by(year) %>%
summarise(sst = mean(sst, na.rm = T),
sst_anom = mean(sst_anom, na.rm = T),
.groups = "keep") %>%
ungroup() %>%
mutate(yr_as_dtime = as.Date(paste0(year, "-07-02")))
# Global Temperature Anomaly Rates to plot as comparison
global_anoms <- read_csv(paste0(oisst_path, "global_timeseries/global_anoms_1982to2011.csv"))
global_anoms <- mutate(global_anoms, year = year(time))
# Get annual summary
global_summary <- global_anoms %>%
group_by(year) %>%
summarise(sst_anom = mean(sst, na.rm = T), .groups = "keep") %>%
ungroup() %>%
mutate(yr_as_dtime = as.Date(paste0(year, "-07-02")))# Build Regression Equation Labels
lm_all <- lm(sst_anom ~ year,
data = filter(annual_summary, year %in% c(1982:2020))) %>%
coef() %>%
round(3)
lm_15 <- lm(sst_anom ~ year,
data = filter(annual_summary, year %in% c(2006:2020))) %>%
coef() %>%
round(3)
lm_global <- lm(sst_anom ~ year,
data = filter(global_summary, year %in% c(1982:2020))) %>%
coef() %>%
round(3)
# Convert yearly rate to decadal
decade_all <- lm_all['year'] * 10
decade_15 <- lm_15['year'] * 10
decade_global <- lm_global["year"] * 10
# Equation to paste in
eq_all <- paste0("y = ", lm_all['(Intercept)'], " + ", decade_all, " x")
eq_15 <- paste0("y = ", lm_15['(Intercept)'], " + ", decade_15, " x")
eq_global <- paste0("y = ", lm_15['(Intercept)'], " + ", decade_global, " x")#### Annual Trend Plot ####
ggplot(data = annual_summary, aes(yr_as_dtime, sst_anom)) +
# Add daily data
geom_line(data = region_timeseries,
aes(time, sst_anom),
alpha = 0.5, color = "gray") +
# Overlay yearly means
geom_line(alpha = 0.7) +
geom_point(alpha = 0.7) +
# Add regression lines
geom_smooth(data = filter(global_summary, year <= 2020),
method = "lm",
aes(color = "1982-2020 Global Trend"),
formula = y ~ x, se = F,
linetype = 3) +
geom_smooth(data = filter(annual_summary, year <= 2020),
method = "lm",
aes(color = "1982-2020 Regional Trend"),
formula = y ~ x, se = F,
linetype = 2) +
geom_smooth(data = filter(annual_summary, year %in% c(2006:2020)),
method = "lm",
aes(color = "2006-2020 Regional Trend"),
formula = y ~ x, se = F,
linetype = 2) +
# Manually add equations so they show yearly not daily coeff
geom_text(data = data.frame(),
aes(label = eq_all, x = min(region_timeseries$time), y = Inf),
hjust = 0, vjust = 2, color = gmri_cols("gmri blue")) +
geom_text(data = data.frame(),
aes(label = eq_15, x = min(region_timeseries$time), y = Inf),
hjust = 0, vjust = 3.5, color = gmri_cols("orange")) +
geom_text(data = data.frame(),
aes(label = eq_global, x = min(region_timeseries$time), y = Inf),
hjust = 0, vjust = 5, color = gmri_cols("green")) +
# Colors
scale_color_manual(values = c(
"1982-2020 Regional Trend" = as.character(gmri_cols("gmri blue")),
"2006-2020 Regional Trend" = as.character(gmri_cols("orange")),
"1982-2020 Global Trend" = as.character(gmri_cols("green")))) +
# labels + theme
labs(x = "",
y = expression("Sea Surface Temperature Anomaly"~degree~C),
caption = paste0("Regression coefficients reflect decadal change in sea surface temperature.
Anomalies calculated using 1982-2011 reference period.")) +
theme(legend.title = element_blank(),
legend.position = c(0.85, 0.1),
legend.background = element_rect(fill = "transparent"),
legend.key = element_rect(fill = "transparent", color = "transparent"),
panel.grid = element_blank())
Quarterly
# Repeat for the seasons (equal quarters here*)
quarter_summary <- region_timeseries %>%
mutate(year = year(time),
season = factor(quarter(time, fiscal_start = 1)),
season = fct_recode(season,
c("Jan 1 - March 31" = "1"),
c("Apr 1 - Jun 30" = "2"),
c("Jul 1 - Sep 30" = "3"),
c("Oct 1 - Dec 31" = "4"))) %>%
group_by(year, season) %>%
summarise(sst = mean(sst, na.rm = T),
sst_anom = mean(sst_anom, na.rm = T),
.groups = "keep")
# Plot
quarter_summary %>%
ggplot(aes(year, sst_anom)) +
geom_line(group = 1) +
geom_point() +
geom_smooth(method = "lm",
aes(color = "Regional Trend"),
formula = y ~ x, se = F, linetype = 2) +
stat_poly_eq(formula = y ~ x,
color = gmri_cols("gmri blue"),
aes(label = paste(..eq.label.., ..rr.label.., sep = "~~~")),
parse = T) +
scale_color_manual(values = c("Regional Trend" = as.character(gmri_cols("orange")))) +
labs(x = "",
y = expression("Sea Surface Temperature Anomaly"~degree~C),
caption = "Regression coefficients reflect annual change in sea surface temperature.") +
theme(legend.title = element_blank(),
legend.position = c(0.875, 0.05),
legend.background = element_rect(fill = "transparent"),
legend.key = element_rect(fill = "transparent", color = "transparent")) +
facet_wrap(~season)
Marine Heatwaves
For the figures below heatwave events were determined using the methods of Hobday et al. 2016 and implemented using the R package {heatwaveR}
The {heatwaveR} package provides a relatively quick way of working with tabular data to calculate a seasonal climate mean, as well as heatwaves and coldwaves at a desired threshold.
These functions can be used to get the climatology using the standard day of year if desired, and track the number and length of heatwave events. Heatwave events follow the definition of Hobday et al. 2016 and are set up to use a 90% threshold for heatwave/coldwave detection.
A marine heatwave is defined a when seawater temperatures exceed a seasonally-varying threshold (usually the 90th percentile) for at least 5 consecutive days. Successive heatwaves with gaps of 2 days or less are considered part of the same event.
The heatwave history for the region displayed above is as follows:
# Use function to process heatwave data for plotting
region_heatwaves <- pull_heatwave_events(region_timeseries, threshold = 90) %>%
distinct(time, .keep_all = T)Heatwave timelines can be then be plotted using the {ggplot2} package for static plots, or by using the {plotly} package for interactive content.
Interactive SST Timeline
For anything we wish to host on the web there is an option to display tables and graphs that are interactive. The {plotly} package is one-such tool for producing plots that allow users to pan, zoom, and highlight discrete observations.
# Grab data from the most recent year through present day to plot
last_year <- Sys.Date() - 365
last_year <- last_year - yday(last_year) + 1
last_yr_heatwaves <- region_heatwaves %>%
filter(time >= last_year)
# Get number of heatwave events and total heatwave days for last year
# How many heatwave events:
last_full_yr <- last_yr_heatwaves %>%
filter(year(time) == year(last_year))
num_events <- max(last_full_yr$mhw_event_no, na.rm = T) - min(last_full_yr$mhw_event_no, na.rm = T)
# How many heatwave days
num_hw_days <- sum(last_full_yr$mhw_event, na.rm = T)# Plot the interactive timeseries
last_yr_heatwaves %>%
filter(time >= last_year) %>%
plotly_mhw_plots()Static SST Timeline
# Set colors by name
color_vals <- c(
"Sea Surface Temperature" = "royalblue",
"Heatwave Event" = "darkred",
"Cold Spell Event" = "lightblue",
"MHW Threshold" = "coral3",
#"MHW Threshold" = "gray30",
"MCS Threshold" = "skyblue",
#"MCS Threshold" = "gray30",
"Daily Climatology" = "gray30")
# Set the label with degree symbol
ylab <- expression("Sea Surface Temperature"~degree~C)
# Plot the last 365 days
ggplot(last_yr_heatwaves, aes(x = time)) +
geom_segment(aes(x = time, xend = time, y = seas, yend = sst),
color = "royalblue", alpha = 0.25) +
geom_segment(aes(x = time, xend = time, y = mhw_thresh, yend = hwe),
color = "darkred", alpha = 0.25) +
geom_line(aes(y = sst, color = "Sea Surface Temperature")) +
geom_line(aes(y = hwe, color = "Heatwave Event")) +
geom_line(aes(y = cse, color = "Cold Spell Event")) +
geom_line(aes(y = mhw_thresh, color = "MHW Threshold"), lty = 3, size = .5) +
geom_line(aes(y = mcs_thresh, color = "MCS Threshold"), lty = 3, size = .5) +
geom_line(aes(y = seas, color = "Daily Climatology"), lty = 2, size = .5) +
scale_color_manual(values = color_vals) +
scale_x_date(date_labels = "%b", date_breaks = "1 month") +
theme(legend.title = element_blank(),
legend.position = "bottom") +
labs(x = "",
y = ylab,
caption = paste0("X-axis start date: ", last_year,
"\nNumber of Heatwave events in ",
year(last_year),
": ",
num_events,
"\nNumber of Heatwave Days in ",
year(last_year),
": ",
num_hw_days,
"\nClimate reference period : 1982-2011"))
Heatwave Events
# Prep the legend title
guide_lab <- expression("Sea Surface Temperature Anomaly"~degree~C)
# Set new axis dimensions, y = year, x = day within year
# use a flate_date so that they don't stair step
base_date <- as.Date("2000-01-01")
grid_data <- region_heatwaves %>%
mutate(year = year(time),
yday = yday(time),
flat_date = as.Date(yday-1, origin = base_date))
# Set palette limits to center it on 0 with scale_fill_distiller
limit <- max(abs(grid_data$sst_anom)) *c(-1,1)
# Assemble heatmap plot
heatwave_heatmap <- ggplot(grid_data, aes(x = flat_date, y = year)) +
# background box fill for missing dates
geom_rect(xmin = base_date, xmax = base_date + 365,
ymin = min(grid_data$year) - .5,
ymax = max(grid_data$year) + .5,
fill = "gray75",
color = "transparent") +
# tile for sst colors
geom_tile(aes(fill = sst_anom)) +
# points for heatwave events
geom_point(data = filter(grid_data, mhw_event == TRUE),
aes(x = flat_date, y = year), size = .25) +
scale_x_date(date_labels = "%b", date_breaks = "1 month", expand = c(0,0)) +
scale_y_continuous(limits = c(1980.5, 2021.5), expand = c(0,0)) +
labs(x = "",
y = "",
"\nClimate reference period : 1982-2011") +
#scale_fill_gradient2(low = "blue", mid = "white", high = "red") +
scale_fill_distiller(palette = "RdBu", na.value = "transparent",
limit = limit) +
#5 inches is default rmarkdown height for barheight
guides("fill" = guide_colorbar(title = guide_lab,
title.position = "right",
title.hjust = 0.5,
barheight = unit(4.8, "inches"),
frame.colour = "black",
ticks.colour = "black")) +
theme_classic() +
theme(legend.title = element_text(angle = 90))
#### summary side plots:
# number of heatwaves
# remove NA as a distinct heatwave number
n_waves <- grid_data %>%
group_by(year(time), mhw_event_no) %>%
summarise(total_days = sum(mhw_event, na.rm = T), .groups = "keep") %>%
ungroup() %>%
drop_na() %>%
group_by(`year(time)`) %>%
summarise(num_waves = n_distinct(mhw_event_no), .groups = "keep")
# average heatwave duration
wave_len <- grid_data %>%
group_by(year(time), mhw_event_no) %>%
summarise(total_days = sum(mhw_event, na.rm = T), .groups = "keep") %>%
ungroup() %>%
drop_na() %>%
group_by(`year(time)`) %>%
summarise(avg_length = mean(total_days), .groups = "keep")
wave_summary <- left_join(n_waves, wave_len)
# Assemble pieces
heatwave_heatmap
Sea Surface Temperature Anomaly Maps
The 2020 global sea surface temperature anomalies have been loaded and displayed below to visualize how different areas of the ocean experience swings in temperature.
# Access information to netcdf on box
nc_year <- "2020"
anom_path <- str_c(oisst_path, "annual_anomalies/1982to2011_climatology/daily_anoms_", nc_year, ".nc")
# Load 2020 as stack
anoms_2020 <- stack(anom_path)Mean Global
The following code will subset the anomalies for July and plot the average sea surface temperature anomalies for that month:
# Get the mean temperature anomalies for July
july_dates <- which(str_sub(names(anoms_2020), 7, 8) == "07")
july_avg <- mean(anoms_2020[[july_dates]])
# Convert wgs84 raster to stars array
july_st <- st_as_stars(rotate(july_avg))
# Set palette limits to center it on 0 with scale_fill_distiller
limit <- max(abs(july_st[[1]]), na.rm = T) * c(-1,1)
# Plot global map
ggplot() +
geom_stars(data = july_st) +
geom_sf(data = world, fill = "gray90") +
scale_fill_distiller(palette = "RdBu", na.value = "transparent", limit = limit) +
map_theme +
coord_sf(expand = FALSE) +
guides("fill" = guide_colorbar(title = "Average Sea Surface Temperature Anomaly",
title.position = "top",
title.hjust = 0.5,
barwidth = unit(4, "in"),
frame.colour = "black",
ticks.colour = "black"))
Regional Map
Using the July average sst anomalies we can then clip the data to just the Caribbean Sea for an aerial view of the region. For this execution the bounding box of lat-lon coordinates from the regional extent will be used.
# Clip Raster - Convert to stars
shape_extent <- c(crop_x, crop_y)
region_ras <- crop(rotate(july_avg), extent(shape_extent))
region_st <- st_as_stars(region_ras)
# Get crop bounds for coord_sf
crop_x <- st_bbox(region_st)[c(1,3)]
crop_y <- st_bbox(region_st)[c(2,4)]
# Zoom out some for cpr extent to be the same as Andy's GOM
if(tolower(params$region) == "cpr gulf of maine"){
crop_x <- c(-71, -65.5)
crop_y <- c(40.5, 45)}
# Set palette limits to center it on 0 with scale_fill_distiller
limit <- max(abs(region_st[[1]]), na.rm = T) * c(-1,1)
# Plot Region - WGS84 Projection
ggplot() +
geom_stars(data = region_st) +
geom_sf(data = new_england, fill = "gray90") +
geom_sf(data = canada, fill = "gray90") +
geom_sf(data = greenland, fill = "gray90") +
scale_fill_distiller(palette = "RdBu", na.value = "transparent", limit = limit) +
map_theme +
coord_sf(xlim = crop_x, ylim = crop_y, expand = T) +
guides("fill" = guide_colorbar(
title = "Average Sea Surface Temperature Anomaly",
title.position = "top",
title.hjust = 0.5,
barwidth = unit(4, "in"),
frame.colour = "black",
ticks.colour = "black"))
Heatwave Progression
Currently in development. Idea is to animate heatwave through the most recent heatwave event.
# Pull the dates of the most recent heatwave
last_event <- max(region_heatwaves$mhw_event_no, na.rm = T)
last_event_dates <- region_heatwaves %>%
filter(mhw_event_no == last_event) %>%
pull(time)
# Buffer the dates
event_start <- (min(last_event_dates) - 7)
event_stop <- max(last_event_dates)
date_seq <- seq.Date(from = event_start,
to = event_stop,
by = 1)
# Load the heatwave dates
data_window <- data.frame(time = c(min(date_seq) , max(date_seq) ),
lon = crop_x,
lat = crop_y)
# Pull data
hw_stack <- oisst_window_load(oisst_path = oisst_path,
data_window = data_window,
anomalies = T)
#drop any empty years that bug in
hw_stack <- hw_stack[map(hw_stack, class) != "character"]##### Format the layers and loop through the maps ####
# Grab only current year, format dates
this_yr <- stack(hw_stack)
day_count <- length(names(this_yr))
day_labs <- str_replace_all(names(this_yr), "[.]","-")
day_labs <- str_replace_all(day_labs, "X", "")
day_count <- c(1:day_count) %>% setNames(day_labs)
# Set palette limits to center it on 0 with scale_fill_distiller
limit <- c(max(values(this_yr), na.rm = T) * -1,
max(values(this_yr), na.rm = T) )
# Plot Heatwave 1 day at a time
day_plots <- imap(day_count, function(date_index, date_label) {
# grab dates
heatwaves_st <- st_as_stars(this_yr[[date_index]])
# plot every day
day_plot <- ggplot() +
geom_stars(data = heatwaves_st) +
geom_sf(data = new_england, fill = "gray90") +
geom_sf(data = canada, fill = "gray90") +
geom_sf(data = greenland, fill = "gray90") +
scale_fill_distiller(palette = "RdBu",
na.value = "transparent", limit = limit
) +
map_theme +
coord_sf(xlim = crop_x, ylim = crop_y, expand = T) +
guides("fill" = guide_colorbar(
title = expression("Sea Surface Temperature Anomaly"~degree~C),
title.position = "top",
title.hjust = 0.5,
barwidth = unit(4, "in"),
frame.colour = "black",
ticks.colour = "black")) +
labs(subtitle = paste("Heatwave Period", event_start, "-", event_stop, "\nDisplay Date:", date_label))
return(day_plot)
})
walk(day_plots, print)